Sampling-Based Sensitivity Analysis of the Mechanistic–Empirical Pavement Design Guide Applied to Material Inputs

Abstract

The Mechanistic–Empirical Pavement Design Guide (MEPDG) is the result of NCHRP Project 1-37A. This pavement design procedure uses mechanistic and empirical models calibrated with national data. Calibration of MEPDG is required to improve the accuracy of the models for local conditions. The data needed for calibration require numerous tests to characterize materials. The fieldwork for collecting the data to verify the models is laborious. This research studied the sensitivity of MEPDG to material input parameters. The main goal of this research was to determine the influence of material parameters on pavement performance as predicted by MEPDG. Results from sensitivity analyses can be used for planning data collection for calibration, implementation, and general understanding of MEPDG. However, MEPDG is so complex that the sensitivity analysis methodology was carefully designed to identify the relative importance of input variables. This research employed space-filling computer experiments with Latin hypercube sampling, standardized regression coefficients, and Gaussian stochastic processes to categorize the relative importance of material inputs in MEPDG for two flexible pavement structures that used Level 3 analysis. The methodology worked well for analyzing the sensitivity of material inputs in MEPDG, with the advantages of varying all parameters at once and requiring a relatively small number of runs for a completed analysis of the entire input space. Effective binder content, as-built air voids, Poisson's ratio, surface shortwave absorption of asphalt layers, and resilient modulus of subgrade had a significant effect on pavement performance.